These data come from the 2016 CCES and allow interested students to model the individual correlates of the Trump vote in 2016. Code/analysis heavily indebted to a 2017 analysis I did on my blog (see references).

## Usage

TV16

## Format

A data frame with 64600 observations on the following 21 variables.

uid

a numeric vector, a unique identifier for the respondent as they first appear in the CCES data.

state

a character vector for the state in which the respondent resides

votetrump

a numeric that equals 1 if the respondent voted says s/he voted for Trump in 2016.

age

a numeric vector for age that is roughly calculated as 2016 - birthyr, as it's coded in the CCES data.

female

a numeric that equals 1 if the respondent is a woman

collegeed

a numeric vector that equals 1 if the respondent says s/he has a college degree

racef

a character vector for the race of the respondent

famincr

a numeric vector for the respondent's household income. Ranges from 1 (Less than $10,000) to 12 ($150,000 or more).

ideo

a numeric vector for the respondent's ideology on a liberal-conservative discrete scale. 1 = very liberal. 5 = very conservative.

pid7na

a numeric vector for the respondent's partisanship on the familiar 1-7 scale. 1 = Strong Democrat. 7 = Strong Republican. Other party supporters (e.g. libertarians) are coded as NA.

bornagain

a numeric vector for whether the respondent self-identifies as a born-again Christian.

religimp

a numeric vector for the importance of religion to the respondent. 1 = not at all important. 4 = very important.

churchatd

a numeric vector for the extent of church attendance for the respondent. 1 = never. 6 = more than once a week.

prayerfreq

a numeric vector for the frequency of prayer for the respondent. 1 = never. 7 = several times a day.

angryracism

a numeric vector for how angry the respondent is that racism exists. 1 = strongly agree (i.e. is angry racism exists). 5 = strongly disagree.

whiteadv

a numeric vector for agreement with statement that white people have advantages over others in the U.S. 1 = strongly agree. 5 = strongly disagree.

fearraces

a numeric vector for agreement with statement that the respondent fears other races. 1 = strongly disagree. 5 = strongly agree.

racerare

a numeric vector for agreement with statement that racism is rare in the U.S. 1 = strongly disagree. 5 = strongly agree.

lrelig

a numeric vector that serves as a latent estimate for religiosity from the bornagain, religimp, churchatd, and prayerfreq variables. Higher values = more religiosity.

lcograc

a numeric vector that serves as a latent estimate for cognitive racism. This is derived from the racerare and whiteadv variables.

lemprac

a numeric vector that serves as a latent estimate for empathetic racism. This is derived from the fearraces and angryracism variables.

## Source

Cooperative Congressional Election Study, 2016

## Details

The latent estimates for religiosity, cognitive racism, and empathetic racism come from a graded response model estimated in mirt. The concepts of "cognitive racism" and "empathetic racism" come from DeSante and Smith.